An Enhanced Real-Time Object Detection Method using Liquid Neural Network and Echo State Network Architecture

Recent developments in Artificial Intelligence (AI) have significantly enhanced autonomous cars' object recognition capabilities, especially with the implementation of deep neural networks. Despite these advancements, achieving a harmonious balance between high precision and speed in vehicle contexts remains a considerable challenge. In this research, a novel object detection model is proposed to locate and identify objects in real-time videos using Liquid Neural Networks (LNNs) variant and Echo State Network (ESN) for model testing and training. One unique feature of LNNs is its capability to dynamically adjust their underlying mechanisms in response to constant streams of fresh data inputs, contributing to their adaptability in dynamic environments. In the analysis, the methodology of the process is presented, comparing LNNs with Artificial Neural Networks (ANNs) and highlighting the superiority of LNNs in the context of object identification for autonomous cars.

I. Introduction

Autonomous car concepts have been around for about eight decades. The development of reliable, long-lasting sensors that are always getting smaller and less expensive, along with recent breakthroughs in wireless connection and communication networks, has been important in the progress of autonomous driving systems. [10] An essential part of these systems is AI. Autonomous vehicles powered by AI must always travel safely in a variety of scenarios. Accurate localization, discrete data collecting, [2] integrated dataset development, and smooth high-level communication with other vehicles and the surrounding intelligent infrastructure are critical to the precision of autonomous navigation. In the future, it's expected that self-driving [13] technology will include buses, mining trucks, freight vehicles, and tractor-trailers.